“EviMass”: A Literature Evidence-Based Miner for Human Microbial Associations

The importance of understanding microbe–microbe as well as microbe–disease associations is one of the key thrust areas in human microbiome research. High-throughput metagenomic and transcriptomic projects have fueled discovery of a number of new microbial associations. Consequently, a plethora of information is being added routinely to biomedical literature, thereby contributing toward enhancing our knowledge on microbial associations. In this communication, we present a tool called “EviMass” (Evidence based mining of human Microbial Associations), which can assist biologists to validate their predicted hypotheses from new microbiome studies. Users can interactively query the processed back-end database for microbe–microbe and disease–microbe associations. The EviMass tool can also be used to upload microbial association networks generated from a human “disease–control” microbiome study and validate the associations from biomedical literature. Additionally, a list of differentially abundant microbes for the corresponding disease can be queried in the tool for reported evidences. The results are presented as graphical plots, tabulated summary, and other evidence statistics. EviMass is a comprehensive platform and is expected to enable microbiome researchers not only in mining microbial associations, but also enriching a new research hypothesis. The tool is available free for academic use at https://web.rniapps.net/evimass.

The selected genera along with its associated genera are displayed as a graph with the central node as the 'query genus' and connected peripheral nodes as the 'associated genera'. The size of the node (and labels) are proportionally scaled to the reported number of associations Click here to view a bar plot version of the network Search and filter the PMID result tables for a specific keyword Export the list in multiple formats  Most frequent words in the PMID result table can be viewed as a 'word cloud' Click (or search from the dropdown) on a node label to view the list of abstracts citing the selected association as well as view the association evidence statistics calculated using hyper-geometric tests Select and add one/more diseases and use the 'Intersection' radio button to filter and view only those associations present in a selected set of diseases Select and add one/more diseases and use the 'Union' radio button to view all associations present in a selected set of diseases Search for a disease name using the searchable dropdown with autosuggest feature

Module 2a
The selected disease along with its associated genera are displayed as a graph with the central node as the 'query disease' and connected peripheral nodes as the 'associated genera'. The size of the node (and labels) are proportionally scaled to the reported number of associations Click (or search from the dropdown) on a node label (genera name) to view the list of other diseases associated with it in form of a bar plot. Additionally the list of abstracts reporting the evidence for the association is also populated.

Nodes in pink represents the taxa identified to be significantly associated with the disease
Clicking on the 'View as Histogram' button displays a bar plot for the selected disease and genera associations. The bar plot is sorted by the number of article evidences for the selected association.
Click (or search from the dropdown)on a node label (genera name) to view the list of other diseases associated with it in form of a bar plot. Additionally the list of abstracts reporting the evidence for the association is also populated.

Bars in pink represents the taxa identified to be significantly associated with the disease
Click (or search from the dropdown)on a node label (genera name) to view the list of abstracts reporting the evidence for the association Search and filter the abstract POS tags for a specific keyword  Secondary search can be performed on the individual columns using the 'column search boxes' Export the list in multiple formats  The most frequently occurring words in a displayed table can be shown as a 'word cloud' as well as a frequency list of top 10 words  A new 'word cloud' can be generated by filtering the table for a custom query Upload a tab delimited edge list file (see sample file 'Download example data') for microbial associations. Node names should be microbial genera names (as in RDP or Green genes database) The uploaded microbial association network will be displayed showing only the edges whose information is available in the EviMass backend database. An edge can be manually selected (by mouse left click) or by using the dropdown search. Once an edge is selected, a tabulated evidence summary for the associations are displayed. The microbe names in the table are hyper linked to their individual association networks.

The edge thickness is mapped proportional to the evidence count observed for the microbe-microbe association
In addition to the tabulated evidence indices, the list of abstracts reporting the selected association are also populated as a separate table below the plot. This table contains only the important keywords from each abstract which can be searched (and filtered) to list only the abstracts containing the desired keyword using the search boxes. The table can be explored in multiple formats and the most frequent words can be viewed as a 'word cloud'.